Causal ECGNet: leveraging causal inference for robust ECG classification in cardiac disorders

Electrocardiogram (ECG) is a graphical representation of the electrical activity of the heart and plays a crucial role in diagnosing heart disease and assessing cardiac function. In the context of infectious diseases, ECG classification is particularly critical, as many infections, such as viral myocarditis and sepsis, can cause significant cardiac complications. Early detection of infection-induced cardiac abnormalities through ECG can provide timely intervention and improve patient outcomes. However, current ECG processing methods often overlook the impact of confounding factors caused by statistical associations, which can compromise classification accuracy, especially in infection-related cardiac conditions. To address this, we propose an innovative approach to causal reasoning based on attention mechanisms. By employing backdoor adjustment for each cardiac lead, our method effectively eliminates confounding factors and models the true causal relationship between ECG patterns and underlying cardiac abnormalities caused by infectious diseases. Furthermore, our approach integrates the concept of entropy with causal inference to enhance ECG classification. By quantifying the information content and variability in ECG signals, we can better identify patterns and anomalies associated with infection-induced cardiac conditions. Experimental results demonstrate that our method achieves significant improvements in classification accuracy and robustness across four benchmark ECG datasets, outperforming existing methods. This work provides a novel perspective on the interplay between infection and cardiac function, offering valuable insights into the detection and understanding of infection-related cardiac complications.